7
Oxysterols act as promiscuous ligands of class-A GPCRs: In silico molecular modeling and in vitro validation Cristina Sensi a , Simona Daniele c , Chiara Parravicini b , Elisa Zappelli c , Vincenzo Russo d , Maria L. Trincavelli c , Claudia Martini c , Maria P. Abbracchio b , Ivano Eberini a, a Laboratorio di Biochimica e Biosica Computazionale, Dipartimento di Scienze Farmacologiche e Biomolecolari, Università degli Studi di Milano, Milan, Italy b Laboratorio di Farmacologia Molecolare e Cellulare della Trasmissione Purinergica, Dipartimento di Scienze Farmacologiche e Biomolecolari, Università degli Studi di Milano, Milan, Italy c Dipartimento di Psichiatria, Neurobiologia, Farmacologia e Biotecnologie, Università degli Studi di Pisa, Pisa, Italy d Cancer Gene Therapy Unit, Program of Immunology and Bio Immuno Gene Therapy of Cancer, Division of Molecular Oncology Scientic, Institute San Raffaele, Milan, Italy abstract article info Article history: Received 15 July 2014 Accepted 15 August 2014 Available online 22 August 2014 Keywords: Oxysterols Cell signaling G proteins Cholesterol Receptors/seven transmembrane domain Molecular recognition According to classical pharmacology, each neurotransmitter/hormonal receptor, including GPCRs, is exclusively activated by highly specic ligands. However, recent evidence challenges this dogma. Oxysterols are produced at inammatory sites and can surprisingly potently activate the Epstein Barr virus induced gene receptor-2 (EBI2), a GPCR involved in adaptive immune responses. Similarly, oxysterols promiscuously operate CXCR2, a chemokine receptor participating to immune reactions and cancer development. Both EBI2 and CXCR2 are phylogenetically related to GPR17, another GPCR implicated in inammatory/immune neurodegenerative events. Here, we used an integrated approach combining comparative modeling, molecular docking and in vitro experiments to investigate their potential interactions with oxysterols. All three receptors share the binding site to allocate oxysterols with different local arrangements, higher sensitivity to specic oxysterols and different activation thresholds. Such differences may dictate the diverse biological effects induced by oxysterols, depending on production site, concentration, specic spatiotemporal features and receptor expression on targeted cells. Thus, EBI2, CXCR2 and GPR17 are promiscuously operated by oxysterols making this class of ligands a l rougelinking oxidative stress, inammation and neurodegeneration. Such a transversal role may represent a conserved, unspecic(but selective) signaling mode, by which emergency molecules activate multiple receptors involved in inammatory/immune responses. © 2014 Elsevier Inc. All rights reserved. 1. Introduction G-protein-coupled receptors (GPCRs) are integral cell membrane receptors organized in seven transmembrane α-helices associated to heterotrimeric G-proteins. They are involved in a plethora of physiolog- ical processes through a very efcient, specic and selective control of cell functions. In several pathologies, a dysregulation of their expression and/or activity has been described, and their pharmacological targeting is at the basis of the most up-to-date therapeutic strategies for many relevant human diseases. Indeed, most marketed drugs have been developed for their ability to target GPCRs, preferentially the class-A ones [1]. In the last decades, classical in vitro and in vivo studies suggested that each GPCR has a specic pharmacological prole and is operated by highly specic ligands. All the pharmacological efforts of the last years have been focused on the development of drugs selective for a single class of GPCR, or for a specic receptor, in order to achieve a successful therapeutic strategy without serious and limiting side effects. However, recent evidence challenges the currently accepted dogma that each receptor responds to a single endogenous ligand or a single family of related signaling molecules. In this respect, oxysterols are well-known as natural and specic ligands of nuclear liver X receptor (LXR) α and β belonging to the cyto- plasmic family of steroid receptors, whose activation controls lipid and cholesterol homeostasis inducing several target gene products. Further- more, activation of LXRs by oxysterols can favor tumor progression by dampening native immune response [2] via activation of specic target genes and trans-repression of pro-inammatory genes. In 2011, Nature published two letters, in which it was demonstrated that EBI2, an orphan class-A GPCR involved in the immune response, can be operated by oxysterols [3,4]. Very recently, our research group contributed to demonstrating the ability of oxysterols to also operate a completely different class-A GPCR, CXCR2, a chemokine receptor involved in the control of the immune system and of cancer develop- ment [5]. All together, these observations suggest that a second level of operability, less specic and more transversal, exists, at least for some class-A GPCRs and for some ligands. The different afnity of Cellular Signalling 26 (2014) 26142620 Corresponding author at: Via Giuseppe Balzaretti, 9, 20133 Milan, Italy. Tel.: +39 02 50318362; fax: +39 02 50318284. http://dx.doi.org/10.1016/j.cellsig.2014.08.003 0898-6568/© 2014 Elsevier Inc. All rights reserved. Contents lists available at ScienceDirect Cellular Signalling journal homepage: www.elsevier.com/locate/cellsig

Oxysterols act as promiscuous ligands of class-A GPCRs: In silico molecular modeling and in vitro validation

  • Upload
    ivano

  • View
    212

  • Download
    0

Embed Size (px)

Citation preview

Cellular Signalling 26 (2014) 2614–2620

Contents lists available at ScienceDirect

Cellular Signalling

j ourna l homepage: www.e lsev ie r .com/ locate /ce l l s ig

Oxysterols act as promiscuous ligands of class-A GPCRs: In silicomolecular modeling and in vitro validation

Cristina Sensi a, Simona Daniele c, Chiara Parravicini b, Elisa Zappelli c, Vincenzo Russo d, Maria L. Trincavelli c,Claudia Martini c, Maria P. Abbracchio b, Ivano Eberini a,⁎a Laboratorio di Biochimica e Biofisica Computazionale, Dipartimento di Scienze Farmacologiche e Biomolecolari, Università degli Studi di Milano, Milan, Italyb Laboratorio di Farmacologia Molecolare e Cellulare della Trasmissione Purinergica, Dipartimento di Scienze Farmacologiche e Biomolecolari, Università degli Studi di Milano, Milan, Italyc Dipartimento di Psichiatria, Neurobiologia, Farmacologia e Biotecnologie, Università degli Studi di Pisa, Pisa, Italyd Cancer Gene Therapy Unit, Program of Immunology and Bio Immuno Gene Therapy of Cancer, Division of Molecular Oncology Scientific, Institute San Raffaele, Milan, Italy

⁎ Corresponding author at: Via Giuseppe Balzaretti, 9, 250318362; fax: +39 02 50318284.

http://dx.doi.org/10.1016/j.cellsig.2014.08.0030898-6568/© 2014 Elsevier Inc. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 15 July 2014Accepted 15 August 2014Available online 22 August 2014

Keywords:OxysterolsCell signalingG proteinsCholesterolReceptors/seven transmembrane domainMolecular recognition

According to classical pharmacology, each neurotransmitter/hormonal receptor, including GPCRs, is exclusivelyactivated by highly specific ligands. However, recent evidence challenges this dogma. Oxysterols are producedat inflammatory sites and can surprisingly potently activate the Epstein Barr virus induced gene receptor-2(EBI2), a GPCR involved in adaptive immune responses. Similarly, oxysterols promiscuously operate CXCR2, achemokine receptor participating to immune reactions and cancer development. Both EBI2 and CXCR2 arephylogenetically related to GPR17, another GPCR implicated in inflammatory/immune neurodegenerativeevents. Here, we used an integrated approach combining comparative modeling, molecular docking andin vitro experiments to investigate their potential interactions with oxysterols. All three receptors share thebinding site to allocate oxysterols with different local arrangements, higher sensitivity to specific oxysterolsand different activation thresholds. Such differences may dictate the diverse biological effects induced byoxysterols, depending on production site, concentration, specific spatiotemporal features and receptorexpression on targeted cells. Thus, EBI2, CXCR2 and GPR17 are promiscuously operated by oxysterols makingthis class of ligands a ‘fil rouge’ linking oxidative stress, inflammation and neurodegeneration. Such a transversalrole may represent a conserved, “unspecific” (but selective) signaling mode, by which emergency moleculesactivate multiple receptors involved in inflammatory/immune responses.

© 2014 Elsevier Inc. All rights reserved.

1. Introduction

G-protein-coupled receptors (GPCRs) are integral cell membranereceptors organized in seven transmembrane α-helices associated toheterotrimeric G-proteins. They are involved in a plethora of physiolog-ical processes through a very efficient, specific and selective control ofcell functions. In several pathologies, a dysregulation of their expressionand/or activity has been described, and their pharmacological targetingis at the basis of the most up-to-date therapeutic strategies for manyrelevant human diseases. Indeed, most marketed drugs have beendeveloped for their ability to target GPCRs, preferentially the class-Aones [1].

In the last decades, classical in vitro and in vivo studies suggested thateach GPCR has a specific pharmacological profile and is operated byhighly specific ligands. All the pharmacological efforts of the last yearshave been focused on the development of drugs selective for a singleclass of GPCR, or for a specific receptor, in order to achieve a successful

0133 Milan, Italy. Tel.: +39 02

therapeutic strategy without serious and limiting side effects. However,recent evidence challenges the currently accepted dogma that eachreceptor responds to a single endogenous ligand or a single family ofrelated signaling molecules.

In this respect, oxysterols are well-known as natural and specificligands of nuclear liver X receptor (LXR) α and β belonging to the cyto-plasmic family of steroid receptors, whose activation controls lipid andcholesterol homeostasis inducing several target gene products. Further-more, activation of LXRs by oxysterols can favor tumor progression bydampening native immune response [2] via activation of specific targetgenes and trans-repression of pro-inflammatory genes.

In 2011, Nature published two letters, in which it was demonstratedthat EBI2, an orphan class-A GPCR involved in the immune response,can be operated by oxysterols [3,4]. Very recently, our research groupcontributed to demonstrating the ability of oxysterols to also operate acompletely different class-A GPCR, CXCR2, a chemokine receptorinvolved in the control of the immune system and of cancer develop-ment [5]. All together, these observations suggest that a second levelof operability, less specific and more transversal, exists, at least forsome class-A GPCRs and for some ligands. The different affinity of

2615C. Sensi et al. / Cellular Signalling 26 (2014) 2614–2620

oxysterolswith respect to CXCR2 endogenous ligands also suggests thatthis second activationmode can become of great interest under specificpathological conditions, e.g. when oxysterols are produced locally atvery high concentrations. These results also raise the hypothesis thatactivation of some class-A GPCRs by oxysterols represents a new wayto modulate some specific immune system function.

At the moment, no molecular data about the activation mechanismof CXCR2 by oxysterols have been reported. In order to carefully analyzeat an atomistic level this mechanism, and to evaluate if oxysterols canact as transversal ligands also for other structurally related class-AGPCRs, we modeled, through comparative in silico modeling, the 3Dstructures of human CXCR2 and of two phylogenetically and structural-ly related receptors, EBI2 and GPR17, a P2Y-like receptor involved incentral nervous system function [6,7]. We then investigated themecha-nism of molecular recognition between these three receptors andoxysterols through in silico molecular docking and well-establishedin vitro experimental approaches. Results show that oxysterols bindto and activate transversally the three receptors, thus extending thenumber of GPCRs and tissues involved in this new type of signaling,with significant implications for the biology and pharmacology ofthese systems.

2. Materials and methods

2.1. Comparative modeling

The tridimensional structures of the human forms of G-proteincoupled receptor 183 (EBI2), C-X-C chemokine receptor type 2(CXCR2), and uracil nucleotide/cysteinyl leukotriene receptor(GPR17) were built through a classical homology/comparativemodeling, by using the MOE Homology Model program of the MOEProtein module of the Molecular Operating Environment (MOE2012.10, Chemical Computing Group, Montreal, Canada). We select-ed, among the human class-A GPCRs, the primary structures whichpresent two conserved pairs of cysteines involved in putative disul-fide bridges: one linking the extracellular loop (EL) 2 to transmem-brane helix (TM) 3, the other connecting N-terminus with EL3. Thissubgroup includes members of the P2Y, CysLT, AGT, and chemokine(CCR, CXCR, CCRL, XCR, CX3CR) receptor families together with APJ,EBI2, GP174, GPR4, GP132, SPR1, G109A, G109B, GP171, GPR31,GPR34, GPR81, P2RY5, P2RY9, and P2Y10 receptors. The primarystructures of the three target proteins were obtained from theUniProt database, and are identified, respectively, with the followingcodes: P32249 for EBI2, P25025 for CXCR2 and Q13304 for GPR17.

Fig. 1. Comparison between CXCR4 and CXCR1 extracellular loops 2. Alpha sphere highlights tcolored according GPCR characteristics. EL2 that overhang the binding site of CXCR4 (A) and o

For all these sequences a global alignment based on T-Coffee algo-rithm [8] has been produced and set as reference alignment for allthe homology modeling procedures. The template for the modelingprocedures of all the selected targets was the crystallographic struc-ture of CXCR4 deposited in the RCSB PDB, and identified as 3ODU [9].The recent NMR structure of the human C-X-C chemokine receptortype 1 (CXCR1) [10], which has a greater identity with respect toCXCR2 than CXCR4, is not a suitable modeling template, since itsorthosteric ligand binding site is blocked by the extracellular loop 2(EL2) (Fig. 1). The ‘automatic disulfide bond detection’ option wasactivated, but the presence and the geometry of the conserved cyste-ines were manually checked. For each receptor, ten models of themainchain were built, and, for each one of these mainchains, onesidechain model was built using the unary quadratic optimization(UQO) procedure [11]. Each single model was submitted to a briefseries of energy minimizations (EMs) meant only to relieve stericstrain and scored according to the GB/IV scoring function [12]. Thetop scoring model for each receptor was further refined through anEM procedure to an RMS gradient value of 0.5 kcal/mol Å. In all themolecular mechanics procedures, the Amber12:EHT forcefield withthe reaction field solvation model. The disulfide bonds not automat-ically detected by the modeling procedure were manually createdthrough the MOE Builder module. The ELs were then submitted toEM runs, after fixing transmembrane helices and intracellular loops(ILs). Six EM runs, all down to an RMS gradient of 0.5 kcal/mol Å,were carried out while restraining the EL atoms with a quadratic forcefrom 105 down to 10−1 kcal/mol Å2. A further EM run was carried outwithout any restraint down to an RMS gradient of 0.5 kcal/mol Å.

The quality of the final model was carefully checked with theGeometry program of the MOE Protein module [13,14].

2.2. Binding site analysis

The binding sites for the three investigated receptors were identifiedthrough the Site Finder program of the MOE Compute module, whichuses a geometric approach to calculate possible binding sites in a recep-tor starting from its 3D atomic coordinates.

2.3. Oxysterol database preparation

Themolecular database file containing all the oxysterols reported inliterature to bind class-A GPCRs was drawn through the MOE builder[3–5,15]. All the structures were energy minimized with the Amber12:

he ligand binding site of both receptors; secondary structures are rendered as ribbon andccupies that of CXCR1 (B) is colored in silver.

2616 C. Sensi et al. / Cellular Signalling 26 (2014) 2614–2620

EHT forcefield in order to produce a single low-energy conformation foreach molecule to test.

2.4. Molecular docking

The in silico accurate docking of the oxysterol database on EBI2,CXCR2, and GPR17 was carried out with the Dock program of the MOECompute module. The docking protocol selected was ‘Rigid Receptor’,with a Triangle Matcher placement methodology, in which the posesare generated by superposing triplets of ligand atoms and triplets ofreceptor site points. The receptor site points are alpha sphere centersthat represent locations of tight packing. The placement procedurereturned 10,000 poses. They were all scored according to the LondondG scoring, which estimates the free energy of binding of the ligandfrom a given pose.

ΔG ¼ cþ Eflex þX

h−bonds

cHB f HB þXm−lig

cM f M þX

atoms i

ΔDi ð1Þ

where c represents the average gain/loss of rotational and translationalentropy; Eflex is the energy due to the loss of flexibility of the ligand(calculated from ligand topology only); fHB measures geometric imper-fections of hydrogen-bonds and takes a value in [0,1]; cHB is the energyof an ideal hydrogen-bond; fM measures geometric imperfections ofmetal ligations and takes a value in [0,1]; cM is the energy of an idealmetal ligation; andDi is the desolvation energy of atom i. The differencein desolvation energies is calculated according to the formula

ΔDi ¼ ciR3i ∭

u∉A∪Buj j−6du−∭

u∉B

uj j−6du

( )ð2Þ

where A and B are the protein and/or ligand volumes with atom ibelonging to volume B; Ri is the solvation radius of atom i (taken asthe OPLS-AA van der Waals sigma parameter plus 0.5 Å); and ci is thedesolvation coefficient of atom i. The coefficients (c, cHB, cM, ci) havebeen fitted from approx. 400 X-ray crystal structures of protein–ligandcomplexes with available experimental pKi data. Atoms are categorizedinto about a dozen atom types for the assignment of the ci coefficients.The triple integrals are approximated using Generalized Born integralformulas. After having scored all the 1000 generated poses, duplicatecomplexes were removed. Poses are considered as duplicates if thesame set of ligand::receptor atom pairs are involved in hydrogen bondinteractions and the same set of ligand atom receptor residue pairs areinvolved in hydrophobic interactions. Only the 30 top scoring solutionswere kept and submitted to a further refinement step, based on energyminimization (EM). In order to speed up the calculation, a grid is usedfor electrostatic calculations during the minimization process. Thedistance-dependent dielectric model was used. The final energy wasevaluated using the GBVI/WSA dG, a forcefield-based scoring functionthat estimates the free energy of binding of the ligand from a givenpose. It has been trained using the MMFF94x and Amber99 forcefieldon the 99 protein–ligand complexes of the SIE training set [16]. Itsfunction form is a linear sum of terms:

ΔG≈cþ α23

ΔEcoul þ ΔEsolð Þ þ ΔEvdw þ βΔSAweighted

� �ð3Þ

where c represents the average gain/loss of rotational and translationalentropy; α and β are constants which were determined during training(alongwith c) and are forcefield-dependent; Ecoul is the coulombic elec-trostatic term which is calculated using currently loaded charges, usinga constant dielectric of 1; Esol is the solvation electrostatic termwhich iscalculated using the GB/VI solvation model; Evdw is the van der Waalscontribution to binding, and SAweighted is the surface area weighted byexposure. This weighting scheme penalizes exposed surface area. All

the 100 refined final poses, after duplicate removal, were stored in theoutput file. All the oxysterols contained in the database were evaluatedaccording the above procedure.

The estimated binding affinity of the top-scoring solution for eachcomplex (receptor::oxysterol) was not directly computed from theGBVI/WSA dG value, but the complexes were further refined throughthe use of a set of specific MOE procedures, named LigX, aimed at theminimization of ligands (oxysterols) in the receptor binding site. ThepKi was computed through the binding free energy estimated with theGBVI/WSA dG scoring function, after complex optimization with LigX.During molecular docking procedures, the Amber12:EHT force fieldwas used both for proteins and oxysterols.

2.5. [35S]GTPγS binding assay

1321N1 pc-DNA 3.1 (control) and HA-Tag GPR17 transfected [17]cells were used in pharmacological assay. Control and transfected cellswere homogenized in 5 mM Tris–HCl-2 mM EDTA (pH 7.4) and centri-fuged at 48,000 g for 15min at 4 °C. The resulting pellets (plasmamem-branes) were washed in 50 mM Tris–HCl and 10 mM MgCl2 (pH 7.4)and stored at −80 °C until used. Compound-stimulated [35S]GTPγSbinding assays in cell-membranes of cells expressing hGPR17 orpcDNA 3.1 were performed as described previously [18]. To determinethe possible interaction between oxysterols and nucleotides, mem-branes were pre-incubated with UDP-glucose (10 μM), and then stimu-lated with different oxysterol concentrations. In some experiments,effects of GPR17 antagonists on compound-mediated effects were eval-uated. In particular, the adenosine-based P2Y12-13 antagonistcangrelor (formerly AR-C69931MX) [19,20], which has been demon-strated to have high affinity for GPR17 [18,21], was used. Different con-centrations of antagonist were added for 10 min before addition of afixed oxysterol concentration (corresponding to almost 10 fold overthe EC50 value) to determine the inhibition of the agonist-mediatedG protein activation. For the analysis and graphic presentation of [35S]GTPγS binding data, a nonlinear multipurpose curve fitting computerprogram, Graph-Pad Prism (GraphPad), was used. All data are present-ed as the mean ± SEM of three different experiments.

3. Results and discussion

3.1. Comparative modeling

Improvement in crystallization and X-ray diffraction methods iscurrently unveiling the structure of an increasing number of class-AGPCRs. This is allowing scientists to dispense the only solved templateavailable so far, crystallized bovine rhodopsin (bRh). In one of ourprevious papers, we used a chimeric approach to model the humanGPR17 structure, trying to keep into account the structural informationprovided by all the solved class-A GPCR structures [22]. Despite the rel-atively low sequence identity between target and templates, the modelwe produced was accurate enough to allow the identification of a set ofvery potent GPR17 agonists. The availability of the crystallographicstructure of the human CXCR4 [9] has made it possible to model EBI2,CXCR2, and GPR17 using a single template. On the other hand, themore recent NMR structure of human CXCR1 [10], although more simi-lar to our receptors, was not useful for our modeling procedure: asshown in Fig. 1, the internal binding site is completely hindered byloop EL2 in a closed conformation.

Table S1, in Supplementary data, reports the identity scores for thealignments between EBI2, CXCR2, GPR17, and, respectively, CXCR4,CXCR1. The alignment between CXCR4 and the selected targets(Supplementary data, Fig. S1) was used to produce the final models ofEBI2, CXCR2, and GPR17.

The Ramachandran φ–ψ dihedral plots show a very good quality forthe three computed models (Supplementary data, Fig. S2); sidechainrotamer and nonbonded contact quality show no specific issues.

2617C. Sensi et al. / Cellular Signalling 26 (2014) 2614–2620

Among them, the three receptors show a very low rootmean squaredeviation (RMSD) values: after superposition, the value for alphacarbons is 1.534 Å. The parts of the superposed receptors differing themost are at the end of transmembrane helix (TM) 3, at the very begin-ning of EL2, and around IL3. The intracellular part of class-A GPCRsis involved in the interaction with the intracellular heterotrimericG-protein, and it will not be considered in our study. In contrast, EL2is a relevant motif for receptor binding site accessibility and for ligandmolecular recognition mechanism [23].

Table 1 lists the amino acids involved in themolecular recognition ofthe ligands for each modeled receptor, and the properties of the identi-fied binding sites. The same table also shows the properties and theamino acids identified by searching for a common space, defined asthe spatial intersection of the volumes of the individual binding sites.

Fig. S3, in Supplementary data, reports the alignment of thesequences of EBI2, CXCR2, and GPR17, in which the amino acids facingthe spatial intersection are highlighted. Some of these amino acids areconserved among the three receptors, defining a molecular basis foroxysterol binding. Furthermore, the differences can help to explain thedifferent affinity for the same oxysterols across the three receptors. Ina previous paper, oxysterol binding to EBI2 was characterized at amolecular level [15]. The authors demonstrated that the mutation toAla and/or Phe of four key residues, namely Arg87, Tyr112, Tyr116and Tyr260, results in a severe decrease in both agonist binding andreceptor activation. The alignment of GPR17, EBI2, and CXCR2, reportedin Supplementary data, Fig. S3, highlights the residues facing into thespatial intersection of the receptor binding sites, in detail: the conservedresidues among the three primary structures (in blue), the partiallyconserved residues (in green), the non-conserved residues (in yellow),and the essential residues for oxysterol-triggered EBI2 activity accord-ing to Ref. [15] (in red). The identified residues in all the three investi-gated receptors are the likely foundation of the molecular basis of thetransversal interaction with oxysterols. The only residue, among thefour identified by Benned-Jensen et al. [15] as critical for oxysterol bind-ing, not directly exposed in the common binding site of the receptors isthe amino acid 3.37 (according to Ballesteros and Weinstein indexingsystem [24]), which in EBI2 can be identified with Tyr116 (see Supple-mentary data, Fig. S4). Residue 3.37 plays several key roles in class-AGPCRs: i) in rhodopsin and in some other class-A GPCRs, the residuein this position belongs to the binding site, and is involved in the ligandmolecular recognition mechanism; ii) in rhodopsin, it is implicated inreceptor motion transduction between TM3 and TM5 [25–27]; and iii)

Table 1Binding site characteristics for each modeled receptor and for the common site. Size indicatesscore for the contact residues. Hyd indicates the number of hydrophobic contact atoms in the r

Receptor Size PLB Hyd Side Residues

EBI2 299 3.93 88 175 ASP20 CYS21 ASP22 LEU23 TYR24 ALA25 TYR38 PHGLU175 ALA176 GLU177 ARG178 THR180 CYS181 MALA200 CYS201 PHE257 TYR260 HIS261 ILE264 HISHIS284 GLN287 ILE288 LEU290 HIS291 VAL294 MET

CXCR2 353 2.94 77 155 PRO38 CYS39 GLU40 PRO41 GLU42 SER43 LEU44 ASSER123 LYS126 GLU127 PHE130 TYR131 VAL180 ARASN203 TRP207 ARG208 MET209 LEU211 ARG212 PILE292 ASP293 ARG294 LEU296 ASP297 GLU300

GPR17 338 3.22 105 184 ASP20 CYS21 ASP22 LEU23 TYR24 ALA25 TYR38 PHGLU175 ALA176 GLU177 ARG178 THR180 CYS181 MALA200 CYS201 PHE257 TYR260 HIS261 ILE264 HISHIS284 GLN287 ILE288 LEU290 HIS291 VAL294 MET

Commonspace

236 2.93 114 256 EBI2LEU23 TYR38 ARG87 TYR90 TYR91 THR107 ALA108TYR260 ILE264 HIS267 LYS271 ARG283 GLN287 LEUCXCR2PRO41 TYR55 TRP104 LYS108 VAL122 SER123 LYS1ARG212 PRO215 TYR267 ASP274 ARG289 ASP293 LGPR17TYR38 ARG87 TYR90 PHE92 PHE111 TYR112 VAL18GLN272 ASN279 ARG280 SER283

in the crystallographic structure of CXCR4, a chemokine receptor,residue 3.37 (Tyr121) is not included in the binding site, and its sidechain is oriented towards TM5, specifically towards His203. As expect-ed, since our EBI2 model was based on CXCR4, residue 3.37 does notbelong to the receptor binding site, and is oriented towards TM5.

3.2. Molecular docking

A set of oxysterols was selected according to their previous use insome experimental papers [3–5] and in experiments carried out inour laboratory. We decided to extend our analysis to EBl2 as well fortwo reasons: i) this receptor was the first class-A GPCR to havebeen demonstrated to be sensitive to oxysterols, including 7α,25-dihydroxycholesterol [3,4,15] and ii) already available data on themolecular characterization of oxysterol binding to EBI2 [15] could bevery useful for the comparison and validation of our molecular model-ing and docking procedures.

Table 2 reports the results of the molecular docking carried outon EBI2. Only the binding free energy of the top-scoring pose for eachtested oxysterol is reported.

All the tested oxysterols showed large negative docking scores(from −9.34 to −8.67 kcal/mol), suggesting that all the tested ligandscan bind to EBI2. 7α,25-Dihydroxycholesterol, with a binding freeenergy value of −9.3 kcal/mol, is the most affine ligand, as alreadyreported both in experimental and computational papers [3,4,15]. Forthe complex EBI2::7α,25-dihydroxycholesterol, we computed, throughLigX, an approx. binding free energy of−10.27 kcal/mol, as described inthe Materials and Methods section. Fig. 2, panel A reports a schematicrepresentation of the interactions of 7α,25-dihydroxycholesterol inthe EBI2 binding site.

Our molecular docking results do not completely agree with previ-ous literature data: only approx. half of the interacting residues arecommon to our and previous docking data, and the overall bindingmode is quite different. We verified if our docking results were biasedby the definition of a bound binding site common to the three investi-gated receptors, but, even when releasing restraints, we could still notreproduce the binding mode reported by Benned-Jensen et al. [15].Specifically, we modified our docking protocol by i) not limiting thebinding site, ii) changing the pose refinement method, and iii) leavingthe side chains in the EBI2 binding site free to rearrange during therefinement procedure. In all cases, our top-scoring solution was alwaysthe same as reported in Fig. 2, panel A. Only the analysis of all the

the number of alpha spheres comprising the site. PLB is the Propensity for Ligand Bindingeceptor. Side indicates the number of sidechain contact atoms in the receptor.

E80 ARG87 TYR90 TYR91 CYS104 THR107 ALA108 PHE111 TYR112 THR115 PRO170ET182 GLU183 TYR184 PRO185 GLU188 GLU189 THR190 PRO194 ILE196 LEU197267 MET268 LYS270 LYS271 ARG273 PHE277 LEU278 GLU279 CYS280 ARG283297 ASN298N47 LYS48 TYR55 TRP104 SER107 LYS108 VAL109 ASN110 GLY111 TRP112 VAL122G184 SER189 VAL192 SER193 PRO194 ALA195 CYS196 TYR197 GLU198 MET200RO215 TYR267 LEU271 ASP274 MET277 ARG278 GLN283 GLU284 THR285 ARG289

E80 ARG87 TYR90 TYR91 CYS104 THR107 ALA108 PHE111 TYR112 THR115 PRO170ET182 GLU183 TYR184 PRO185 GLU188 GLU189 THR190 PRO194 ILE196 LEU197267 MET268 LYS270 LYS271 ARG273 PHE277 LEU278 GLU279 CYS280 ARG283297 ASN298

PHE111 TYR112 THR180 CYS181 MET182 GLU183 THR190 PRO194 ILE196 LEU197290 HIS291 VAL294 CYS295

26 ARG184 CYS196 TYR197 GLU198 MET200 ASN203 ARG208 MET209 LEU211EU296 ASP297 GLU300

0 CYS181 LEU182 GLN183 TYR185 LYS188 ALA189 HIS192 TYR251 TYR258 TYR262

Table 2Physico-chemical parameters for the tested oxysterols.

Receptor Oxysterol Docking score[kcal/mol]

Affinity[kcal/mol]

EBI2 7α,25-Dihydroxycholesterol −9.34 −10.2727-Hydroxycholesterol −8.98 −9.1022R-Hydroxycholesterol −8.68 −9.5925-Hydroxycholesterol −8.67 −9.64

CXCR2 22R-Hydroxycholesterol −8.20 −9.797α,25-Dihydroxycholesterol −8.13 −8.43

GPR17 27-Hydroxycholesterol −8.61 −7.8022R-Hydroxycholesterol −8.10 −10.147α-Hydroxycholesterol −7.69 −8.40

2618 C. Sensi et al. / Cellular Signalling 26 (2014) 2614–2620

30 produced solutions allowed the identification of a pose similar to theone proposed by Benned-Jensen et al. for 7α,25-dihydroxycholesterol[15], however its docking score had a very unfavorable value of−4.30 kcal/mol. Furthermore, according to the same authors, Tyr116plays a key role in the formation of a hydrogen bond with the 3β-OHof the ligand, but in our EBI2 model, as in the CXCR4 template, theorientation of the side chain of Tyr 3.37 is outward from the bindingsite, as extensively discussed in some lines above. Literature experi-mental data [3,4] demonstrated that other oxysterols such as25-hydroxycholesterol are less potent agonist of EBI2. Also fromour simulations, 25-hydroxycholesterol has an approx. binding freeenergy of −9.64 kcal/mol, suggesting that its affinity for EBI2 islower than the affinity of 7α,25-dihydroxycholesterol.

We have previously demonstrated that oxysterols are able to bindthe human chemokine receptor in vitro and activate its signal transduc-tion pathways [5], but we have not yet described their binding mode.No atomic data have been published about the molecular recognitionmechanism between CXCR2 and these ligands. CXCR2 is a receptor forpeptides, and its binding site is wider than that of the other class-AGPCRs that recognize small ligands. Usually the binding site of theclass-AGPCRs,which bind peptides, is organized in two distinct regions:an external site I which is involved in the initial peptide-receptor recog-nition, and an inner site II, in which the peptide triggers the receptorresponse [9]. As it can be observed from the published crystallographicstructure of CXCR4, small organic ligands exert their activity by bindingto the orthosteric site II. From our docking results, the tested oxysterolsseem to be able to bind CXCR2 in its orthosteric binding site II, with adocking score ranging between−8.20 and−8.13 kcal/mol. We have al-ready demonstrated that 22R-hydroxycholesterol is an agonist, able tobind the orthosteric site II of CXCR2. In fact, in a displacement experiment,we observed that increasing concentrations of 22R-hydroxycholesterolare able to compete with 125I-IL-8, when binding to CXCR2, suggesting

Fig. 2. Binding fashion of: (A) 7α,25-dihydroxycholesterol in the EBI2 binding site; (B)hydroxycholesterol and cangrelor in the GPR17 binding site. Receptor secondary structures aare rendered as stick.

that both ligands compete for one (at least partially) common bindingsite. The evidence that 22R-hydroxycholesterol cannot completelydisplace 125I-IL-8 from CXCR2 confirms that oxysterols, and specifically22R-hydroxycholesterol, can only compete for the orthosteric bindingsite II, while not altering the weaker interaction between 125I-IL-8 andCXCR2 in its binding site I. Fig. 2, panel B shows the binding mode ofthe complex CXCR2::22R-hydroxycholesterol, in which binding site II isefficiently occupied by the tested ligand.

We also computed the dissociation constant of the simulatedcomplex CXCR2::22R-hydroxycholesterol, in order to compare it withour previously published experimental results. An approx. pKi value of6.6 was obtained, which corresponds to a dissociation constant (Ki) ofapprox. 0.27 μM. This value is in good agreementwith the experimentalEC50, 1.32 μM, obtained for the same complex. It is well known that EC50equals the sum of the Ki values as well as one half the total concen-tration of target receptor [28]. The docking scores for oxysterols ofEBI2 and CXCR2 fall within a comparable energy range. Differencesare however detectable with specific molecules. For instance,EBI2::7α,25-dihydroxycholesterol has an approx. binding free ener-gy of −10.26 kcal/mol, whereas CXCR2::22R-hydroxycholesterol isassociated with the lower value of −9.79 kcal/mol. These resultssuggest that the interactions of the selected receptors withoxysterols are significantly different, suggesting a varying sensitivityand selectivity of this class of ligands vs their target receptors.

Our group has a long experience on GPR17 that we cloned anddeorphanized some years ago as a dual uracil nucleotide/cysteinyl-leukotriene receptor [18]. The phylogenetic proximity between GPR17and CXCR2/EBI2 let us hypothesize that the activation mechanismmediated in the latter by oxysterols could be shared by GPR17, suggest-ing a transversal operability of these receptors mediated by unconven-tional ligands. For this reason, we investigated this possibility throughmolecular modeling, docking and in vitro experiments.

In silico GPR17 showed a high affinity for the tested oxysterols, withdocking scores ranging from −8.61 to −7.69 kcal/mol. As alreadydiscussed in the ‘Comparative modeling’ section, a portion of theoxysterol binding site of GPR17 is shared with the other investigatedclass-A GPCRs. Furthermore, a comparison between the binding modeof the potent already identified orthosteric and the top-scoring dockingpose on GPR17 of the most potent oxysterol suggests that 22R-hydroxycholesterol efficiently binds to the already identified orthostericbinding site. All the three tested oxysterols showahigh binding free ener-gy, with the following affinity order: 22R-hydroxycholesterol N 7α-hydroxycholesterol N 27-hydroxycholesterol, as reported in Table 2. Forinstance, according to our experimental data, 22R-hydroxycholesterol isthe most potent among the tested oxysterols on both CXCR2 andGPR17: its EC50 for CXCR2 is 1.32 μM, whereas the same value for

22R-hydroxycholesterol in the CXCR2 binding site; and (C) superposition of 22R-re rendered as ribbon and colored according GPCR characteristics; chemical compounds

Table 3Pharmacological parameters for the tested compounds.

Compound EC50

[nM]Emax

(% vs basal)IC50

Compound vs cangrelor[nM]

27-Hydroxycholesterol 4.99 ± 0.78 173.5 ± 6.2 0.59 ± 0.067α-Hydroxycholesterol 0.70 ± 0.09 161.8 ± 2.4 0.88 ± 0.0122R-Hydroxycholesterol

0.21 ± 0.03 151.7 ± 3.9 0.89 ± 0.09

2619C. Sensi et al. / Cellular Signalling 26 (2014) 2614–2620

GPR17 is 0.7 nM, which entails a difference in its potency of approx. athousand fold. No experimental data are available about the same li-gand on EBI2, but from our docking simulations, reported in Table 2,7α,25-dihydroxycholesterol shows a better docking score, followedby 27-hydroxycholesterol and 22R-hydroxycholesterol. In the nextsection, these data will be discussed vs experimental evidence ob-tained through a pharmacological reference assay for GPCR activity.As positive control, we also docked cangrelor, a potent orthostericantagonist of the uracil nucleotide binding site, on GPR17 [18,23].From our simulations, cangrelor is predicted to strongly bindGPR17, with a negative docking score of −9.90 kcal/mol and withan approx. binding free energy of −11.59 kcal/mol. Moreover,from a structural point of view, both 22R-hydroxycholesterol andcangrelor bind the same site on GPR17 with a potential competitivebehavior. The high binding free energy computed for cangrelor, associ-ated with its null intrinsic activity, suggests that on GPR17 cangrelorshould behave as a competitive antagonist of oxysterols, and specificallyof 22R-hydroxycholesterol. Fig. 2, panel C depicts the superposition ofthe top-scoring poses of 22R-hydroxycholesterol and cangrelor,

Fig. 3. In vitro results. Dose–response curves of new compounds and effect of UDP-glucose onwith (A) hGPR17 or (B) pc-DNA3.1 transfected 1321N1 cells were incubated with different cthe Materials and methods section. All data are expressed as percentages of basal [35S]GTPγS bduplicate. (C) Effect of the GPR17 receptor antagonist cangrelor on ligand-stimulated [35S]GTwith cangrelor (0.1 nM–100 nM), and then stimulated with a selected concentration for each lin theMaterials and methods section. All data are expressed as percentages of basal [35S]GTPγSduplicate. Membranes from GPR17-expressing 1321N1 cells were incubated with UDP-glucosehydroxycholesterol or (E) 7α-hydroxycholesterol or (F) 22R-hydroxycholesterol. [35S]GTPγS biexpressed as percentages of basal [35S]GTPγS binding (set to 100%) and are the mean (SEM) o

showing their partial overlap and accounting for the competitive antag-onism of cangrelor on oxysterol activity.

3.3. In vitro experiments

Data on the ability of oxysterols to operate EBI2 and CXCR2 havealready been published [3–5]. Conversely, no reports or data areavailable about the effect of oxysterols on GPR17. In order to evaluatethe functional activity on GPR17 of the selected compounds and tovalidate the computational data, the selected oxysterols were testedin a well-established GPCR assay, the [35S]GTPγS binding, based onthe ability of agonists to increase the binding of radioactive GTP tothe activated GPCR. The [35S]GTPγS binding results are summarizedin Table 3.

Dose–response curves of tested compounds obtained by [35S]GTPγSassay in 1321N1 cells expressingGPR17 are shown in Fig. 3. All the com-pounds were able to stimulate GTPγS binding, in a concentration-dependent manner, with affinity constant values in the nanomolar orsubnanomolar range. No significant binding stimulation was observedin 1321N1 pcDNA 3.1 control cells, confirming that the observed effectswere due to a specific GPR17 activation. As predicted by molecularmodeling, all the three tested oxysterols bind to GPR17 and, acting asagonists potently activates it. It is not possible to easily correlate thecomputed approx. binding free energy to the experimental EC50

values, but it is noteworthy that the computationally predicted affin-ity order for the three tested oxysterols is fully consistent with theobtained experimental EC50 values (22R-hydroxycholesterol N 7α-hydroxycholesterol N 27-hydroxycholesterol), as shown in Table 3.In order to confirm our in silico predictions, the modulation of

ligand dose–response curves. Membrane aliquots obtained from 1321N1 cells transfectedompound concentrations, and [35S]GTPγS binding assay was performed as described ininding (set to 100%) and are mean ± SEM of 3 different experiments, each performed inPγS binding. Membranes from hGPR17-transfected cells were pre-incubated for 10 minigand (10 fold over the EC50 value). [35S]GTPγS binding assay was performed as describedbinding (set to 100%) and are mean± SEM of 3 different experiments, each performed in(10 μM), and then stimulated with different concentrations (0.01 nM–10 μM) of (D) 27-nding assaywas performed as described in theMaterials andmethods Section. All data aref 3 different experiments, each performed in duplicate.

2620 C. Sensi et al. / Cellular Signalling 26 (2014) 2614–2620

GTPγS binding stimulated by a high concentration of the tested com-pounds in the presence of different concentrations of cangrelor wasevaluated.

4. Conclusions

Data about the ability of non-conventional ligands to operate class-AGPCRs have been accumulating. Specifically, increasing evidence indi-cates that oxysterols, oxidized derivatives of cholesterol, are involvedin many activities that are not strictly associated with cholesterolmetabolism andmay act as emergency signalingmolecules in neurode-generative disorders, including demyelinating diseases [29]. Differentresearch groups, ours included, have recently described the ability ofoxysterols to bind not only to their canonical receptors (the oxysterolreceptors LXRs), but to also show a ‘promiscuous’ behavior in activating,in a specific but not selective manner, other receptors, namely selectedGPCRs. In particular, the three investigated GPCRs show a commonbinding space, in which oxysterols can place themselves with differentlocal arrangements. This evidence may explain the different potencyof the same oxysterol on each individual receptor. Our data suggestthat each of these three class-A GPCRs shows higher sensitivity to aspecific oxysterol and a different activation threshold which potentiallydifferentiates the final biological effect of these compounds, dependingon the site of production, their concentration, specific spatio-temporalfeatures and the typical receptor expression pattern of the targetedcell/tissue. Besides being common molecular targets of oxysterols,EBI2, CXCR2 and GPR17 are phylogenetically related to each other andalso participate in CNS inflammatory responses. Notably, these threereceptors respond to other already characterized independent familiesof endogenous ligands, which are known to participate to the onsetand/or resolution of the inflammatory reaction. It is not a chance thatoxysterols are also produced locally at high concentrations under dis-tinct pathological conditions, such as ischemic, inflammatory, neurode-generative and neoplastic diseases. Of interest, UDPglucose, one of thecanonical ligands of GPR17, increases the affinity or potency of 22R-,7α- and 27-hydroxycholesterol, suggesting that conventional andnon-conventional ligands may work together under danger conditions.We thus hypothesize that oxysterols may act as immediate emergencysignals alerting specific GPCRs, likely to ‘prime’ towards their conven-tional ligands; alternatively, oxysterols may act transversally to‘synchronize’ some GPCRs and induce them to act together, maybe viathe formation of homomers or heteromersmediating distinct biologicaleffects. Such a transversal role for oxysterols challenges a classicalpharmacological paradigm according to which a family of endogenousligands specifically interacts with just one single class-A GPCR. Accord-ing to our data, CXCR2 and GPR17 are operated by oxysterols as promis-cuous ligands, making this class of ligands a ‘fil rouge’ linking oxidativestress, inflammation and neurodegeneration. Future studies will clarifythe extent of oxysterol involvement in class-A GPCR inter-operabilityand the pathophysiological significance of this common transversalreceptor activation process.

Acknowledgments

Authors are deeply grateful to the Italian Multiple Sclerosis Founda-tion (FISM) for the financial support (Project N. 2013/R1 to MPA).

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.cellsig.2014.08.003.

References

[1] B.C. Heng, D. Aubel, M. Fussenegger, Annu. Rev. Pharmacol. Toxicol. 54 (2014)227–249.

[2] C. Traversari, V. Russo, Curr. Opin. Pharmacol. 12 (2012) 729–735.[3] S. Hannedouche, J. Zhang, T.S. Yi, W.J. Shen, D. Nguyen, J.P. Pereira, D. Guerini,

B.U. Baumgarten, S. Roggo, B. Wen, R. Knochenmuss, S. Noel, F. Gessier, L.M. Kelly,M. Vanek, S. Laurent, I. Preuss, C.Miault, I. Christen, R. Karuna,W. Li, D.I. Koo, T. Suply,C. Schmedt, E.C. Peters, R. Falchetto, A. Katopodis, C. Spanka, M.O. Roy, M. Detheux,Y.A. Chen, P.G. Schultz, C.Y. Cho, K. Seuwen, J.G. Cyster, A.W. Sailer, Nature 475(2011) 524–527.

[4] C.L. Liu, X.V. Yang, J.J. Wu, C. Kuei, N.S. Mani, L. Zhang, J.X. Yu, S.W. Sutton, N. Qin,H. Banie, L. Karlsson, S.Q. Sun, T.W. Lovenberg, Nature 475 (2011) 519–523.

[5] L. Raccosta, R. Fontana, D. Maggioni, C. Lanterna, E.J. Villablanca, A. Paniccia,A. Musumeci, E. Chiricozzi, M.L. Trincavelli, S. Daniele, C. Martini, J.A. Gustafsson,C. Doglioni, S.G. Feo, A. Leiva, M.G. Ciampa, L. Mauri, C. Sensi, A. Prinetti, I. Eberini,J.R. Mora, C. Bordignon, K.R. Steffensen, S. Sonnino, S. Sozzani, C. Traversari, V. Russo,J. Exp. Med. 210 (2013) 1711–1728.

[6] C.J. Raport, V.L. Schweickart, D. Chantry, R.L. Eddy Jr., T.B. Shows, R. Godiska, P.W. Gray,J. Leukoc. Biol. 59 (1996) 18–23.

[7] R. Fredriksson, M.C. Lagerstrom, L.G. Lundin, H.B. Schioth, Mol. Pharmacol. 63 (2003)1256–1272.

[8] C. Notredame, D.G. Higgins, J. Heringa, J. Mol. Biol. 302 (2000) 205–217.[9] B.Wu, E.Y. Chien, C.D.Mol, G. Fenalti,W. Liu, V. Katritch, R. Abagyan, A. Brooun, P.Wells,

F.C. Bi, D.J. Hamel, P. Kuhn, T.M. Handel, V. Cherezov, R.C. Stevens, Science 330 (2010)1066–1071.

[10] S.H. Park, B.B. Das, F. Casagrande, Y. Tian, H.J. Nothnagel, M. Chu, H. Kiefer, K. Maier,A.A. De Angelis, F.M. Marassi, S.J. Opella, Nature 491 (2012) 779–783.

[11] P. Labute, Proteins 75 (2009) 187–205.[12] M. Wojciechowski, B. Lesyng, J. Phys. Chem. 108 (2004) 18368–18376.[13] M.J. Bower, F.E. Cohen, R.L. Dunbrack Jr., J. Mol. Biol. 267 (1997) 1268–1282.[14] S.C. Lovell, I.W. Davis, W.B. Arendall III, P.I. de Bakker, J.M. Word, M.G. Prisant,

J.S. Richardson, D.C. Richardson, Proteins 50 (2003) 437–450.[15] T. Benned-Jensen, C. Norn, S. Laurent, C.M.Madsen, H.M. Larsen, K.N. Arfelt, R.M.Wolf,

T. Frimurer, A.W. Sailer, M.M. Rosenkilde, J. Biol. Chem. 287 (2012) 35470–35483.[16] M. Naim, S. Bhat, K.N. Rankin, S. Dennis, S.F. Chowdhury, I. Siddiqi, P. Drabik, T. Sulea,

C.I. Bayly, A. Jakalian, E.O. Purisima, J. Chem. Inf. Model. 47 (2007) 122–133.[17] S. Daniele, M.L. Trincavelli, M. Fumagalli, E. Zappelli, D. Lecca, E. Bonfanti, P. Campiglia,

M.P. Abbracchio, C. Martini, Cell. Signal. 26 (2014) 1310–1325.[18] P. Ciana, M. Fumagalli, M.L. Trincavelli, C. Verderio, P. Rosa, D. Lecca, S. Ferrario,

C. Parravicini, V. Capra, P. Gelosa, U. Guerrini, S. Belcredito, M. Cimino, L. Sironi,E. Tremoli, G.E. Rovati, C. Martini, M.P. Abbracchio, EMBO J. 25 (2006)4615–4627.

[19] F. Marteau, E. Le Poul, D. Communi, D. Communi, C. Labouret, P. Savi, J.M. Boeynaems,N.S. Gonzalez, Mol. Pharmacol. 64 (2003) 104–112.

[20] M. Fumagalli, L. Trincavelli, D. Lecca, C. Martini, P. Ciana, M.P. Abbracchio, Biochem.Pharmacol. 68 (2004) 113–124.

[21] A.M. Pugliese, M.L. Trincavelli, D. Lecca, E. Coppi, M. Fumagalli, S. Ferrario, P. Failli,S. Daniele, C. Martini, F. Pedata, M.P. Abbracchio, Am. J. Physiol. Cell Physiol. 297(2009) C1028–C1040.

[22] I. Eberini, S. Daniele, C. Parravicini, C. Sensi, M.L. Trincavelli, C. Martini, M.P. Abbracchio,J. Comput. Aided Mol. Des. 25 (2011) 743–752.

[23] C. Parravicini, G. Ranghino, M.P. Abbracchio, P. Fantucci, BMC Bioinformatics 9(2008) 263.

[24] J.A. Ballesteros, H. Weinstein, Methods Neurosci. 25 (1995) 62.[25] K. Sansuk, X. Deupi, I.R. Torrecillas, A. Jongejan, S. Nijmeijer, R.A. Bakker, L. Pardo,

R. Leurs, Mol. Pharmacol. 79 (2011) 262–269.[26] J.A. Ballesteros, L. Shi, J.A. Javitch, Mol. Pharmacol. 60 (2001) 1–19.[27] J.A. Goncalves, K. South, S. Ahuja, E. Zaitseva, C.A. Opefi, M. Eilers, R. Vogel, P.J. Reeves,

S.O. Smith, Proc. Natl. Acad. Sci. U. S. A. 107 (2010) 19861–19866.[28] J.R. Griffiths, Biochem. Soc. Trans. 6 (1978) 258–260.[29] C. Garenc, P. Julien, E. Levy, Free Radic. Res. 44 (2010) 47–73.